计算机与数字工程2025，Vol.53Issue(4)：1194-1200,7.DOI:10.3969/j.issn.1672-9722.2025.04.046
基于YOLOv5改进算法在机械零件识别中的研究
Research on Improved Algorithm Based on YOLOv5 in Mechanical Parts Recognition
张浩洋 1何仕荣 1孟冬平 1张倩倩1
作者信息
- 1. 上海理工大学机械工程学院 上海 200093
- 折叠
摘要
Abstract
Aiming at the problems of misidentification and missing identification of mechanical parts during sorting and assem-bly in the industrial production process,a high-precision mechanical parts detection algorithm based on YOLOv5 is proposed.This algorithm is mainly aimed at the low detection accuracy of the original model for small-sized parts,and the detection layer of the de-tection network is designed to increase the detection layer of the detection network to better integrate the semantic information and position information of the parts.In addition,in order to overcome the redundant feature information of the parts extracted by the original model,SE layer is embedded in the feature extraction part of the model to help the new model extract more mechanical part features related to part identification.Finally,experiments are carried out on the ten self-made mechanical parts datasets,and the effectiveness of the algorithm is verified.The results show that the newly designed mechanical parts detection algorithm improves mAP by 1.45%compared with the original YOLOv5 detection algorithm.关键词
深度学习/YOLOv5/注意力机制/特征融合/零件识别Key words
deep learning/YOLOv5/attention mechanism/feature fusion/part recognition分类
信息技术与安全科学引用本文复制引用
张浩洋,何仕荣,孟冬平,张倩倩..基于YOLOv5改进算法在机械零件识别中的研究[J].计算机与数字工程,2025,53(4):1194-1200,7.
